This invention relates to a digital processing system which system is capable of processing large quantities of voice, sonar, radar, video and other types of analog signals and to operate on these signals by performing complicated arithmetic operations in general and more particularly to such a system and architecture to enable the processing of analog signals to provide digital signals which can be employed for narrow band secure communications and voice recognition.
As one can readily perceive, the field of digital signal process (DSP) is extremely complicated, and many systems exist which essentially operate on analog signals to provide bandwidth reduction and to enable digitized encrypted representatives of the analog signal to be sent over conventional communication channels.
The analog signals are operated on by the digital signal processing circuitry by subjecting the signals to algorithms. The algorithms employed can be quite complicated. For example, a technique that is widely employed is referred to as linear predictive coding (LPC). Essentially, LPC is used to model the vocal tract with a multisection all pole filter to synthesize the voice on the receiving end. These filter coefficients are derived by analyzing the voice at the transmitter by several different techniques. One technique employed is the covariance method whereby a 10.times.10 covariance matrix is generated and the filter coefficients are derived by Cholesky decomposition.
In addition to the filter coefficients, the pitch periods of the speech are derived by typically autocorrelation of the digitized voice samples or by an absolute magnitude difference function. As indicated these techniques have been fairly well documented, and for example of the mathematics involved and to be handled by such systems, reference is made to an article entitled SPEECH ANALYSIS AND SYNTHESIS BY LINEAR PREDICTION OF THE SPEECH WAVE by B. S. Atal and S. L. Haneuer, published in the JOURNAL OF THE ACCOUSTICAL SOCIETY OF AMERICA 50,2 (1971) pages 637 to 655.
See also the following articles A LEVEL COMPLETING DYNAMIC TIME WARPING ALGORITHM FOR CONNECTED WORD RECOGNITION by C. S. Myers and L. R. Rabiner published in IEEE TRANSACTIONS ON ACCOUSTIC SPEECH AND SIGNAL PROCESS, Vol. ASSP-29 No. 2,(April 1981). As one can ascertain by these articles and a number of others, the circuitry for operating on analog signals using LPC techniques is extremely complicated and by the use of computers involves extremely large memories which often are implemented by adjunct devices such as magnetic tape, disks and so on.
Apart from the above processing techniques, there is a field of information transfer which is referred to as data encryption. Cryptography is the science of transforming messages for the purpose of making the message unintelligible to all but the intended receiver of the message. The term data encryption refers to the use of cryptographic methods in communications for the same reasons. The transformation used to encipher a message typically involves both a general method, or algorithm and a key. These systems are also extremely complicated in that the purpose of the system is to prevent an unauthorized user from accessing information which may be transmitted along a conventional communication channel.
According to modern day cryptographic techniques, a number of systems are employed which include public-key cryptosystems using complicated ciphers such as the RSA cipher. Other techniques are based on the difficulty of determining which subset of a given list of numbers adds up to a given target number. In these systems, both the enciphering and deciphering of data involves arithmetic operations which are manifested in the use of extremely large digital numbers, that is, numbers having a large number of bits where these numbers may be continuously raised to appropriate powers employing modular arithmetic and other sophisticated mathematical techniques.
See for example an article entitled A METHOD FOR OBTAINING DIGITAL SIGNATURES AND PUBLIC-KEY CRYPTOSYSTEMS by R. Rivest, A. Shamir and L. Adleman, published in COMMUNICATIONS ACM, pages 120 to 126 (February 1978). This is a general treatise concerning such systems and the mathematical approaches. See also an article entitled PRIVACY AND AUTHENTICATION: AN INTRODUCTION TO CRYPTOGRAPHY by W. Diffie and M. Hellman, published in the PROCEEDINGS OF THE IEEE, Vol. 67, (March 1979) pages 397 to 472. Essentially, based on modern communication techniques, it is apparent that these digital signal processing applications impose requirements that exceed the capabilities of any available signal chip signal processor such as those presently commercially available. The prior art, cognizant of these techniques, attempted to solve these problems by using multiple types of chips---such as bit-slice chips, wide memory configurations, multiplier-accumulators and a large quantity of MSI/SSI glue logic. These techniques were implemented in an attempt to build a processor unit having more arithmetic capability. Thus in these approaches various components require multiple, interconnected integrated circuits.
There has been an approach to alleviate the problem in employing circuit element designated as a dynamically programmable processing element (DPPE). A DPPE is employed in a digital signal processing system or as a digital signal processor (DSP) which requires complicated arithmetic procedures to be implemented and hence for use in the systems described above. The DPPE is a special purpose computer which essentially has a program bus for transmitting and receiving program data from an external source. The unit also includes a data bus which bus can transmit or receive digital data. Coupled between the program and data buses are input and output registers or buffers which are capable of storing transmitted or received data propagating on either of the buses.
The unit typically contains a program memory which has an input coupled to the program bus and an output which can also access the program bus. Apart from the program memory, there is an addressable data memory which can be addressed from the program bus and is operative to transmit data to the data bus. Thus the program memory can receive program information or data from an external source as well as receiving data from the data memory which can also transfer data to the external source. Thus the unit can be reprogrammed in real time by the external source and the DPPE can also execute program instructions by fetching them from the external source as well as executing its own program instructions as stored in its program memory.
Hence the structure and format of the DPPE enables it to interface with an external source such as a microprocessor to assist and perform program instructions as well as to perform arithmetic operations on data from the microprocessor and to communicate with the microprocessor after completion of the various routines. For an example of such a device, reference is made to U.S. Pat. No. 4,439,839 entitled DYNAMICALLY PROGRAMMABLE PROCESSING ELEMENT issued Mar. 27, 1984 to K. N. Kneib and G. Vensko and assigned to the International Telephone and Telegraph Corporation, the assignee herein.
In spite of the improvements offered by the DPPE, many signal processing applications exist which exceed the capability of available commercial integrated circuits, including the DPPE.
Thus it is an object of the present invention to provide a multiple processor architecture which is adapted to process digital signals and which employs a plurality of DPPE's under control of a master microprocessor. In regard to such systems, the multiple processor system, according to this invention, can be employed to solve full-duplex LPC voice and modem algorithms as well as to be used to perform data encryption utilizing complicated cipher algorithms. The same architecture can be used for implementing other communications speed and speaker recognition algorithms which require extensive processing in implementing the arithmetic operations.
Thus according to the system architecture to be described, all such complicated digital signal processing techniques can be accommodated by simply adding additional modules to the system architecture of this invention.